Adaptable protocol stack architecture for future sensor networks
Umakishore Ramachandran, Rajnish Kumar · 2006
Current technology trends suggest that futuristic wireless sensor networks (FWSN) are well equipped to support applications such as video surveillance and emergency response that have, in addition to high computation and communication requirements, a need for dynamic data fusion. Because of the dynamic and heterogeneous nature of FWSN environment, adaptability is emerging as a basic need in designing network protocols. Adaptability can be broadly supported at two levels: network level and node level. At the network level, different nodes are the entities that adapt their roles in a cooperative manner to improve network level cost metrics. Similarly, at the node level, different protocol modules are the entities that adapt their behavior in a holistic manner to best perform the roles assigned to the node. The goal of this thesis is to provide an adaptable protocol stack architecture for data fusion applications. Towards this goal, the thesis presents a software architecture called SensorStack, that raises and answers three set of questions. First, towards network-level adaptability, how best to dynamically adapt the placement of a fusion application task graph on the network? How best to support such an adaptation in an application-specific manner? We have designed a distributed role assignment algorithm and implemented it in the context of DFuse, a framework for distributed data fusion. Simulation results show that the role assignment algorithm significantly increases the network lifetime over static placement. Second, towards node-level adaptability, how best to facilitate cross-layering on a node to foster agile adaptation of a node's behavior commensurate with the network-level changes? SensorStack provides information exchange service (IES), a framework for cross-module information exchange. IES can be thought of as a centrally controlled bulletin-board within each node where different modules can post available data, or request information, and get notification when information becomes available. IES preserves the benefits of layering while facilitating adaptability. IES has been implemented in TinyOS and Linux, to show both the feasibility of the design as well as demonstrate the utility of cross-layering to increase application longevity. Third, how best to integrate node-level and network-level adaptability? How best to design a scalable dissemination protocol for sharing information across nodes? Towards tying the network and node level adaptability together, control data published in IES needs to be shared across the network. SensorStack uses a probabilistic broadcast-based dissemination service (IDS) for control data. Simulation experiments show that IDS allows nodes to share IES data more efficiently than using multiple diffusion trees. More broadly, an efficient and reliable dissemination of information over a large area is a critical ability of a sensor network for various reasons such as software updates and transferring large data objects (e.g., camera images). Thus, for bulk data broadcast, we design FBcast, an extension of the probabilistic broadcast based on the principles of modern erasure codes. Simulation results on TOSSIM show that FBcast offers higher reliability with lower number of retransmissions than traditional broadcasts.